DialogLab: Authoring, Simulating, and Testing Dynamic Human-AI Group Conversations
summary
DialogLab is a prototyping toolkit that lets designers build multi-party human‑AI scenes with drag‑and‑drop avatars, attach content, and configure free and moderated party modes plus intra‑party speaking rules; it simulates flow with snippet-based phases and connects snippets into non‑linear transitions. Designers run live replays or a simulated human mode with editable suggestions to stress‑test interruptions and topic shifts, then export logs for exportable verification with turn‑distribution visuals. An evaluation with end users and domain experts shows faster iteration and more structured verification, and the code is being open sourced.
AI-written summary · about
chapters (7)
- 0:00Introduction and motivation
- 2:58Challenges in multi-party conversations
- 4:02Authoring group dynamics
- 7:58Conversation flow dynamics
- 9:46Testing and simulating conversations
- 10:50Applications and future work
- 11:52Conclusion and Q&A
related talks
01h 04mBuilding Expertise Into Generated Conversations
01h 04mBuilding Expertise Into Generated Conversations
51mRolePl-AI
42mConversational Assistants that Converse with Each Other
01h 02m[Full Workshop] Building Conversational AI Agents
10mBuilding Conversational AI Agents at Scale
01h 02m"This is fAIne". Panel discussion
32mParty like it's 1970: Conversational interfaces are back (into your Android app)
